Working paper: Can machine learning models capture correlations in corporate distresses?

Working Paper - October 2018 - No. 128

Authors Mølgaard, Pia; Matin, Rastin
Subject Credit risk; Risk management
Type Working paper
Year 2018
Published 26 October 2018
We implement a regularly top-performing machine learning model and find that the added complexity in the model does not imply that the model is better at capturing correlation in corporate distresses compared to traditional distress models. Instead, we propose a frailty model, which allows for correlations in distresses. This model demonstrates competitive performance in terms of ranking firms by their riskiness, while providing accurate risk measures of a corporate loan portfolio.